Queues Optimization

121.book Seite 197 Freitag, 3.
August 2007 2:19 14
Data sent to CRM is usually written into the inbound queues

first and then further processed by CRM Middleware. It is
not surprising, therefore, that inbound queues are the first
place in CRM Middleware where optimization can begin.
Inbound Queues
If we consider the different systems that can be connected to CRM,

the SAP R/3 system in particular is the one that can cause problems
for CRM due to its large quantities of data. If you change a business
object in R/3, the change is transferred directly to CRM and saved in
an inbound queue. The inbound queue scheduler takes the data from
the queue and transfers it to the R/3 Adapter (see Figure 8.1).
Inbound Processing
Inbound Queues
R/3
R3A*
BAPI
R/3 Adapter
Groupware
ISP*
XML
Groupware
Adapter
Validation
mBDoc
CRM SITE*
sBDoc
CRM SITE*
Mobile
Clients
CRM
Application
Mobile Adapter
Outbound Processing
Groupware
Online
DB
CRM SITE*
mBDoc
Outbound Queues
R/3
Validation Service
CSA*
Replication Service
R3A*
ISP*
BAPI
R/3 Adapter
XML
Groupware
Adapter
p
Synchronization Flow
sBDoc
Mobile Bridge
CRM SITE*
CRM SITE*
Mobile
Clients
Mobile
Adapter
CDB
Service
CDB
CRM SITE*
sBDoc
R&R Service
Figure 8.1 Inbound Queues in CRM Middleware
197
121.book Seite 198 Freitag, 3. August 2007 2:19 14
Inbound Queues
This delta supply of data works well in normal operations. However,

if a connected system sends a large quantity of data faster than CRM
can process it, the data accumulates in the inbound queue, and the
corresponding repercussions, as was explained with examples in
Chapter 7, Introduction to Performance Optimization, can occur if settings are incorrect. Therefore, the objective should be to optimize
the inbound queue processing as much as possible.
All queues in CRM Middleware are based on Remote Function Calls
(RFCs). RFC is therefore critically important for processing in Middleware. Inbound and outbound queues are implemented using
qRFC (queued RFC), although Replication & Realignment queues are
based on tRFC (transactional RFC) rather than qRFC. It is also important to note that, in addition to R3A*, ISP* and CRM_SITE* inbound
queues, CSA* queues are also implemented as inbound queues as of
CRM Release 4.0. Figure 8.2 shows CRM Middleware from the perspective of RFC. For all inbound queues, there is only one inbound
queue scheduler.
Inbound qRFC
Inbound Processing
Inbound Queues
R/3
R3A*
BAPI
R/3 Adapter
Groupware
ISP*
XML
Groupware
Adapter
Validation
mBDoc
CRM
M SITE*
S
sBDoc
CRM
M SITE*
S
Validation Service
CRM
Application
Mobile Adapter
Online
DB
Mobile
Clients
CRM
M SITE*
S
Outbound Processing
Outbound qRFC
R/3
CSA*
Replication Service
R3A*
Groupware
Inbound qRFC
mBDoc
Outbound Queues
ISP*
BAPI
R/3 Adapter
XML
Groupware
Adapter
p
sBDoc
Mobile Bridge
CRM SITE*
CRM SITE*
Mobile
Clients
Mobile
Adapter
CDB
Service
CRM SITE*
sBDoc
R&R Service
CDB
tRFC
Figure 8.2 RFC in CRM Middleware
Chapter 2, Inbound Processing and Validation, contains a detailed

description of inbound queues and qRFC.
198
The Number of Inbound Queues Is Too High
Different reasons can account for why inbound queues are processed
slowly:
So many inbound queues are created that the inbound queue

scheduler experiences problems in processing these queues (see
Section 8.1).
Only a few queues are created, however, these contain a very high
number of data records and the available resources in CRM cannot
be used effectively (see Section 8.2).
All the CRM systems work processes are occupied by the inbound
queue processing and a resource bottleneck occurs, which slows
down the processing (see Section 8.3).
The inbound processing of individual data records is slow (see

Section 8.4).
Dependencies exist between inbound queues (see Section 8.5).
In the following sections, we will describe the different causes of

problems for inbound queue processing and introduce possible solutions.
8.1
The inbound queue scheduler ensures that the inbound queues are
processed in parallel. To do this, it occupies all work processes available to it. However, the increasing number of queues impairs the
performance of the scheduler. The deterioration in performance is
clearly noticeable after 10,000 inbound queues. At worst, the performance is impaired to such a degree that it takes longer to assign the
next queue entry to a free work process than it takes to post the
entry itself. From a performance point of view, this means that the
inbound queues are actually processed sequentially, in other words,
they are no longer processed in parallel, and the available hardware
resources are not used (see example below). The only way to prevent
this is to limit the number of inbound queues in CRM.
199
Causes of
problems
8.1
Inbound Queues
Example
The company, BodeGolzeSchrder Inc. (hereafter abbreviated to BGS),

has 2.7 million customers worldwide, 500,000 of which are in Germany
and are serviced by 500 field sales representatives. BGS performs a
realignment in R/3 once a year. The sales territories of the field sales representatives are reassigned during this realignment. This reorganization
does not affect 40 % of customers (200,000), since they have traditionally
been serviced by the same field sales representative for years and the goal
here is to retain this established customer relationship. The remaining
60 % are assigned the new sales areas.1 BGS has developed two reports
for this purpose. The Z_DELETE_ASSIGNMENT report deletes the
assigned employee responsible for a customer, and the Z_CALCULATE_
NEW_ASSIGNMENT report uses a range of criteria to calculate which
employee will be responsible for the customer in future. Reports Z_
DELETE_ASSIGNMENT and Z_CALCULATE_NEW_ASSIGNMENT change
300,000 data records each. Due to the delta supply of data, 300,000
inbound queues with two entries each are generated in CRM. In addition
to the day-to-day activities, CRM Middleware must also process the
600,000 BUPA_REL data records. In BGSs CRM system, the R/3 Adapter
and validation usually need two seconds to process a BUPA_REL BDoc.
Due to the extremely high number of inbound queues, the inbound
queue scheduler needs 2.2 seconds to assign a queue entry to a work process. This results in an estimated inbound processing duration of 600,000
2.2 = 1.32 million seconds (which is equivalent to 15 days, 6 hours).2
Solution
approaches
Reducing single
records
There are three approaches that you can adopt to reduce the number
of inbound queues: 12
Reduce the single records in R/3
Change the naming for queues
Use R/3 parameter CRM_MAX_QUEUE_NUMBER_DELTA (which

ultimately also affects the naming of queues)
The first approach involves reducing the single records generated by

R/3. The company BGS (from our example) has this option. You
could change the reports in R/3 to the effect that the COMMIT
WORK is executed only after n data records, not after every single
data record, and n objects are written into a BUPA_REL BAPI. If we
pick 500 as the number to represent n, for example, the Z_DELETE_
1 In most cases, the result is the same as before, whereby the field sales representative retains most of his customers.
2 The impact of daily activities and the duration of outbound processing were not
included in this estimate.
200
8.1
ASSIGNMENT report would only create 600, rather than 300,000,

inbound queues. Unfortunately, this option is not always available.
The other factor to bear in mind with this solution is that many (500)
CSA queue entries can be generated from one R3A queue entry.
Since CSA queues are also inbound queues in terms of CRM Middleware, they also overload the inbound queue scheduler.
The second approach consists of limiting the number of inbound
queues by changing the naming for queues. R/3 queues for the delta
supply of data in CRM have the following names by default: R3AD_
<object part><object ID>. There is one entry in the CRMQNAMES
table for each object type. The object part element of the queue
name is located in the QOBJPART field, and the BAPIFLD field indicates which field is used to fill the object ID. Figure 8.3 shows the
entry for the BUPA_REL object. The R/3 outbound queue (and the
CRM inbound queue) would have the queue name R3AD_
BUPA12345 for the object 12345. Chapter 2, Inbound Processing and
Validation, provides a detailed description of the naming conventions for the different inbound queues for delta loads, initial loads,
and so on (R3AD*, R3AI*, etc.).
Figure 8.3 R3AD Queue Naming; Table CRMQNAMES in SAP R/3 (Transaction SE16)
201
Changing the
naming for queues
Inbound Queues
By changing the naming for queues, several business partners are

written into the same queue, rather than each business partner being
written into its own queue. You control the maximum number of
queues that can be generated and which object instances are written
into a particular queue by determining which and how many positions from the object ID are important for the queue name. You use
the LENGTH field to control the number of relevant positions for
naming queues and the FLDOFFSET field to select the position.
For example, if the LENGTH field has the value 1 and FLDOFFSET
has the value 9, this means that all object instances that have the
same number on the tenth position are written into the same queue.
You can therefore create a maximum of 10 queues.
The queue name is determined by the first object ID, for which a
queue is generated, and does not change again until the queue has
been processed completely (and disappears). The left-hand column
of the table in the example below contains the R/3 object ID of the
business partner. The same business partners are used in all four
examples, but the values differ for the LENGTH and FLDOFFSET
fields. The right-hand column of the table contains the name of the
inbound queue, into which the business partner is written. As you
can see from this example, the number of queues depends on the
field values and IDs of the business partners.
Example
1. Setting: Standard Naming

Result: The six customers are written into six queues.
202
R/3 Customer Number
CRM Queue Name
0000054601
R3AD_CUSTOME0000054601
0000034541
0000099421
0000028421
0000028422
0000067822
2. Setting: FLDOFFSET = 9 and LENGTH = 1

Result: The six customers are written into two queues.
R/3 Customer Number
CRM Queue Name
0000054601
0000034541
0000099421
0000028421
0000028422
0000067822
Result: The six customers are written into four queues.
R/3 Customer Number
CRM Queue Name
0000054601
0000034541
0000099421
0000028421
0000028422
0000067822
Result: The six customers are written into five queues.
R/3 Customer Number
CRM Queue Name
0000054601
0000034541
0000099421
0000028421
0000028422
0000067822
203
8.1
Inbound Queues
Changing the
naming for queues
Proceed as follows to change the naming for queues (see Figure 8.3):
1. In the CRMQNAMES table in R/3, find the entry with the object
name OBJNAME, for which you want to change the queue naming
(Transaction SE16, or alternatively, Transaction SM30).
2. Enter the corresponding values for the Field Offset (FLDOFFSET)
and Field Length (LENGTH) parameters.
3. Make sure that the total of the two values does not exceed the
maximum length of the object ID. The maximum length of the
object ID is specified in the BAPIFLDLEN field.
CSA queues
Proceed as follows to change the naming for CSA queues (see Figure
8.4):
1. In the SMOFQFIND table in CRM, find the entry with the object
name in the BDoc Type column, for which you want to change the
queue naming (Transaction SM30, or alternatively, Transaction
SE16).
2. Enter the values for the Field Offset (corresponds to FLDOFFSET)
and Internal Length (corresponds to LENGTH) parameters.
3. The entries for all objects written into the same queue must be
changed; in other words, all entries for which the values in the
fields Queue Object Part and Segment Field are the same.
Note that you may only change the queue naming if the queues are
empty.
Using the CRM_

MAX_QUEUE_
NUMBER_DELTA
parameter
The third approach for reducing the number of inbound queues

requires using the CRM_MAX_QUEUE_NUMBER_DELTA parameter. This parameter is maintained in the CRMPAROLTP and determines how many queues are created, irrespective of the settings in
the CRMQNAMES table. If the parameter is set, the ASCII values of
the last three positions of the queue name are combined and calculated modulo CRM_MAX_QUEUE_NUMBER_DELTA. The result is a
number that is lower than 1,000 and smaller than CRM_MAX_
QUEUE_NUMBER_DELTA. The new queue name is created based on
this number.
You can maintain the number of queues differently for each object
type.3
3 As of PI_BASIS 2006.1, or if you have implemented SAP Note 944633.
204
Figure 8.4 CSA Queue Naming; Table SMOFQFIND in CRM (Transaction SM30)
Proceed as follows to limit the number of R3AD queues (see Figure

8.5):
1. Create a new data record for the CRMPAROLTP table in R/3 (Transaction SM30).
2. Enter the value CRM_MAX_QUEUE_NUMBER_DELTA in the
Parameter name (PARNAME) field.
3. Enter the name of the object type in the Param. Name 2
(PARNAME2) field.
4. Enter the name for your CRM system in the User (CONSUMER)
field.
5. Enter the number of queues in the Param. Value (PARVAL1)
field.
205
Limiting R3AD
queues
8.1
Inbound Queues
Figure 8.5 Table CRMPAROLTP in SAP R/3

Advantages and
disadvantages
When you change the naming for queues, you can use the LENGTH
field to determine the maximum number of queues that can be created, however, you cannot determine how many are actually created. The maximum number of queues is not only influenced by the
value of the LENGTH field, but also by the type of object ID. Table
8.1 illustrates this dependency. However, it also shows that you can
only limit the maximum number of queues in very granular steps.
You cannot limit the maximum number of queues to 25, for example.
Type of Object ID
Value of LENGTH
Numeric
10
100
1,000
36
1,296
46,656
Alphanumeric
Max. Number of Queues
Table 8.1 Maximum Numbers of Inbound Queues Depending on the Type of

Object ID and the LENGTH Parameter
The actual number of queues is influenced to a certain extent by the

value of the FLDOFFSET field. Depending on the selection and status
of the number ranges, certain positions in the object ID have the
same value for all instances, whereas, for others, all values are found.
206
8.1
In contrast to changing the queue name, you can limit the maximum
number of queues exactly by using the CRM_MAX_QUEUE_
NUMBER_DELTA parameter. The type of object ID is not important
in this case. Difficulties only occur if the last three positions of the ID
are not distributed equally. This problem does not occur when you
use the standard naming function (unless there are fewer than 1,000
objects). However, if you have implemented your own rules when
assigning IDs, you may not be able to use the parameter as shown in
the BGS example.
Example
To be able to identify a customers country of origin immediately based on

the customer number, all customer numbers at BGS end with a two-digit
country code (external number assignment). When the CRM_MAX_
QUEUE_NUMBER_DELTA parameter is used, a maximum of 10 queues
are therefore created for the realignment in Germany.
You could say that the CRM_MAX_QUEUE_NUMBER_DELTA

parameter is generally better suited to limiting queues than it is to
changing the queue naming. To achieve optimum processing, it is
worthwhile to distribute the data records across all queues as evenly
as possible.
You may only change the queue naming or the CRM_MAX_QUEUE_
NUMBER_DELTA parameter if all inbound queues are processed.
Otherwise, data inconsistencies may occur between R/3 and CRM, as
illustrated in the following example.
Attention
Example
Before you implement the change, all data records relating to customer A
are written into queue A. After you make the change, they are saved in
queue X. If you implement the change, even though queue A still contains
data, two queues will exist for a certain period of time in the system that
contain data from customer A. Queue A contains the older data records
and queue X contains the newer data records. If queue X is processed
before queue A, the newer data records supersede the older data records.
Consequently, the newer data records are posted in CRM first and then
the values in the older data records overwrite the newer values.
We recommend that you first deregister the outbound queues in R/3

and then wait until the R3AD inbound queues have been processed
completely in CRM. Then make the change to the CSA queue naming
207
Reducing SAP R/3

outbound queues
Inbound Queues
and register the queues again. You will find it harder to change the
naming of R3A queues, because you cannot prevent outbound
queues from being created in R/3. You can therefore only implement
a change within a maintenance window if no data is created in R/3,
which is transferred to CRM.
8.2
The Number of Inbound Queues Is Too Low
In Section 8.1, we described how you could limit the number of

inbound queues. We also mentioned that the actual number of
inbound queues can be a lot smaller and data records are not necessarily replicated evenly across all queues. The following example
highlights the problems that can occur if you limit the number of
queues too much.
Example
BGS has decided not to change the report, but instead has chosen to
change the queue naming. To avoid overloading the CRM system too
much, the value 1 is selected for the LENGTH parameter and the value 7
is chosen for FLDOFFSET. BGSs CRM system consists of an application
server and a database server with four CPUs each. There are 20 dialog
work processes in each case on both servers. The CRM system should easily be able to process 10 inbound queues in parallel. BGS expects that 10
queues with 72,000 data records each will be created and that the processing will take 40 hours, based on 72,000 * 2 seconds = 144,000 seconds. After the mass change is started, it becomes apparent that the longest of the 10 queues contains 151,200 data records (i.e., 21 % of the
changed customers have the number 3 as the last number before the
country code in their customer number; 14 % have the number 8; 10 %
have the numbers 7 and 5; 9 % have the numbers 6 and 9; 8 % have the
numbers 2 and 4; 7 % have the number 0; and 4 % have the number 1).
The processing for this queue takes 3.5 days, based on 151,200 2 seconds = 302,400 seconds.
In some circumstances, you can solve the problem described in the

above example by analyzing the customer numbers and choosing the
corresponding FLDOFFSET parameter. However, this analysis is very
laborious and there is also no guarantee that it gives you an ideal
value for FLDOFFSET. Alternatively, the maximum number of
inbound queues can further increase. The crucial factor here is to
208
Hardware Bottlenecks
8.3
choose the correct balance between the number of inbound queues

that is too low and too high.
8.3
In addition to the number of inbound queues, there are other parameters that you must take into account in order to guarantee optimum
processing of inbound queues. This is demonstrated in the following
example.
Example
For the next test, BGS selects the value 2 for the LENGTH parameter and
the value 6 for FLDOFFSET. After the mass change starts, 84 inbound
queues, rather than the expected 100, are generated in CRM with an
average of 8,500 entries. The inbound queue scheduler uses the available
resources to process the queues as quickly as possible, and occupies all
dialog work processes. This results in an overload situation (CPU bottleneck). The CRM system is still busy processing only the inbound queues
and no other work can be performed with the CRM system (this also
applies in particular for the system administrator). The processing time of
the individual data records also multiplies due to the CPU bottleneck.
The problems that BGS experienced in the last example were not
caused by the number of inbound queues being too high, but due to
the fact that the inbound queue scheduler occupied too many dialog
work processes of the CRM system. The inbound queue scheduler
doesnt have the necessary logic to check how heavily the system is
already being utilized and subsequently to decide whether other
queue entries can be processed, or whether there should be a break
in the processing. All work processes available to the inbound queue
scheduler are used for the processing until all inbound queues are
empty.
You can only prevent this kind of overloading by limiting the number of dialog work processes that the inbound queue scheduler is
allowed to occupy. For this purpose, the inbound queue scheduler is
assigned to an RFC server group. You create and maintain RFC server
groups in Transaction RZ12. Figure 8.6 shows a CRM system with
three RFC server groups. The first group (without a name) is the standard group. This group is always used if there is no explicit assign-
209
RFC server group
Inbound Queues
ment to a group. The other two groups, Queue_Scheduler and

parallel_generators, were created manually and one instance each
was assigned to both groups. In principle, several instances can also
be assigned to a group. In such cases, when a user logs on, the
instance with the best response times is determined automatically
and the user is logged on to this instance.
To create an RFC server group, start Transaction RZ12, and click the
New button (Edit Create assignment menu option). To display the
resource assignment of an instance or to change it, double-click the
corresponding row. A Change Assignment dialog window appears
with the RFC parameter values of the instance.
Figure 8.6 RFC Server Groups (Transaction RZ12)
210
The parameters are used as follows:
Parameters of the
RFC server group
Activated (0 or 1)
Switch for activating the determination of resources. This should

always have the default value 1 (= active).
Max. Requests in Queue (%)
Quote for the number of maximum pending requests in the dialog

waiting queue of the dispatcher, which is proportionate to the
maximum length of the dispatcher request queue. The default
value is 5.
Max. No. of Logons (%)
This value specifies the maximum percentage of logons allowed to

this instance by asynchronous RFCs (the total number of logons is
contained in the rdisp/tm_max_no parameter). The remaining
percentage continues to be reserved for dialog and HTTP users,
i.e., if the number of logons exceeds this value, the caller will not
be assigned any resources. The default value is 90.
Maximum No. Separate Logons (%)
This value specifies the maximum percentage of logons allowed to

this instance by asynchronous RFCs of a user (the total number of
logons is contained in the rdisp/tm_max_no parameter). If the
number of separate logons exceeds this value, the user is not
assigned any further resources. Ideally, this value should not be
greater than the Max. No. of Logons (%) parameter. The default
value is 25.
Max. Number of WPs Used (%)
Quote for the number of dialog work processes that a user is

allowed to use. If the number of dialog work processes used
exceeds this value, the caller is not assigned any further dialog
work processes. This quote prevents all dialog work processes
from being occupied by a users RFCs. The default value is 75.
The system doesnt check the user name, which means that, if a
user logs on to the system several times, each logon is viewed as a
separate user, even though the user name is the same.
8.3
Minimum Number of Free WPs
Quote for the number of dialog work processes that must be

reserved for other users. If the number of free dialog work processes is less than the number specified in the quote, the caller is
not assigned any dialog work processes. The default value is 1.
211
Inbound Queues
The value must always be smaller than the number of dialog work
processes (rdisp/wp_no_dia parameter); otherwise, an RFC
request cannot be processed.
Max. No. of Comm. Entries (%)
Quote for the maximum number of communication entries of an

instance that may be used by parallel RFCs (the total value of the
communication entries is contained in the rdisp/max_comm_
entries parameter). If the number of entries used exceeds this
value, the caller is not assigned any resources. The default value is
90.
Max. wait time
Maximum number in seconds that a work process can remain in

idle mode if it doesnt receive any resources after the load on the
system has been checked. The actual wait time is determined from
the available resources. The fewer resources available, the longer
the wait time.
Profile parameters
of RFC server
group
All settings that you implement in Transaction RZ12 are immediately

active, however, they are only saved up until the next time when you
restart the instance. After you restart the instance, the settings are
lost and the old values are active again. To save the values permanently, you must save them as profile parameters. Table 8.2 contains
the names of the profile parameters for the individual values in
Transaction RZ12.
RZ12
Profile Pparameters
Activated (0 or 1)
rdisp/rfc_use_quotas
Max. Requests in Queue (%)
rdisp/rfc_max_queue
Max. No. of Logons (%)
rdisp/rfc_max_login
Maximum No. of Separate Logons (%)
rdisp/rfc_max_own_login
Max. Number of WPs Used (%)
rdisp/rfc_max_own_used_wp
Minimum Number of Free WPs
rdisp/rfc_min_wait_dia_wp
Max. No. of Comm. Entries (%)
rdisp/rfc_max_comm_entries
Max. wait time
rdisp/rfc_max_wait_time
Table 8.2 Names of Profile Parameters
212
You assign the inbound queue scheduler to an RFC server group in

Transaction SMQR. You select the menu option Edit Change AS
group and enter the name of the RFC server group.
8.3
Assigning an RFC
server group
Figure 8.7 shows that the Name of AS Group (DEFAULT = All)

parameter no longer has the DEFAULT value; it now has the
Queue_Scheduler value instead.
Figure 8.7 Assigning the RFC Server Group (Transaction SMQR)
The optimum setup of RFC server groups depends on the available

hardware, volume of data and scenarios used. If you use a pure
mobile sales scenario, online or Internet sales users must not be considered when the system is being optimized. In this case, almost all
work processes can be made available to the RFC. In all other cases,
you must limit the resources for the RFC in such a way that the users
will still be able to continue using the system, even if the RFC load is
high. If the CRM system has more than one application server, you
may find it useful to set up the RFC server groups in such a way that
the RFC processing is restricted to the resources of one application
server, while the online users use a different application server. By
adopting this approach, you can ensure that the users and the RFC
will not disturb one another. This holds true especially if data is
transferred from the backend, or from/to the mobile clients and the
CRM application is used intensively (i.e., both scenarios occur concurrently). The disadvantage of this setup is that the RFC is restricted
to one server if the RFC load is high (e.g., if you implement a mass
change), even though resources are available on another server (e.g.,
during the night).
213
Optimum
setting for the RFC
server group
Inbound Queues
If there is only one application server, it is critical that you optimize

the parameters of the RFC server group. The distribution of
resources should reflect the actual load distribution RFC load versus online load in each case.
To prevent the inbound queue scheduler from using all work processes of the CRM system, you must set the Max. Number of WPs
Used (%) and Minimum Number of Free WPs parameters correctly.
The Minimum Number of Free WPs parameter determines how
many work processes are available for the RFC and the Max. Number
of WPs Used (%) parameter specifies how many work processes one
RFC user is allowed to use.
If the system has sufficient dialog work processes, the Minimum
Number of Free WPs parameter should have the value 3, rather than
the default value 1. This ensures that you can still log on to this
instance through the SAPGUI, even if the load on the system is very
high. This recommendation applies only to a pure mobile sales scenario or a pure RFC instance. In all other cases, the value should be
greater, and should also be based on the number of simultaneously
active (concurrent) online users.
When determining the Max. Number of WPs Used (%) parameter,
remember that the RFC is used not only by the inbound queue
scheduler, but by other users as well (replication & realignment, outbound queue scheduler, external systems, and especially mobile clients through the DCOM station/.NET Connector). If you select a
value that is too high, although the inbound queues will be processed quickly, the data will accumulate in other areas of the CRM
Middleware because resources will not be available there. You can
only achieve an optimum processing speed if all the systems components have sufficient resources. We discuss these dependencies in
the CRM Middleware in more detail in Chapter 13, Performing Optimized Mass Changes.
Detailed documentation about configuring the system resources for
parallel RFCs, tRFCs, and qRFCs is available under SAP NetWeaver in
the SAP Help Portal (http://help.sap.com). Click Search Documentation. As Search string, enter Configuration of System Resources for
Parallel RFCs tRFC qRFC. Choose SAP NetWeaver and the release
and language that you want, and click Search.
214
Performance Analysis for Processing Single Records
8.4
Performance Analysis for Processing Single

Records
SAP provides an entire range of statistical information and trace options

for a performance analysis. Some of these are listed in Table 8.3.
Transaction
Description
ST03N
Workload Monitor
STAD
Statistical Records
ST12
Single Transaction Trace
SE30
Runtime Analysis (ABAP Trace)
ST04
Database Performance Analysis
DB02
Database Performance: Tables and Indices
ST10
Table Call Statistics
ST05
Performance Analysis (SQL Trace)
SQLR
SQL Trace Interpreter
SMWMFLOW
Message Flow Statistics
SMWT
Middleware Trace
Table 8.3 Performance Analysis Transactions
Except for the Message Flow Statistics (Transaction SMWMFLOW)

and the Middleware Trace (Transaction SMWT), which well discuss
in further detail, the analysis transactions are standard (i.e., not
CRM-specific transactions that are found in every SAP system).
Detailed descriptions about these transactions and the most suitable
way that you can use them are already available; therefore, we dont
want to repeat them here.4
However, we would like to point out how the performance analysis
of CRM Middleware differs from the performance analysis in other
SAP systems. The user who executes a transaction often acts as a filter
for the performance analysis, in order to find the correct statistical
data records or trace a particular action in the system. The user is
not always identified very easily in CRM Middleware. For example,
4 We recommend that you refer to the book, SAP Performance Optimization Guide:
Analyzing and Tuning SAP Systems, by Thomas Schneider, published by SAP PRESS
(4th edition, 2005).
215
Analysis
transactions
8.4
Inbound Queues
you will only be able to clearly establish which user is processing an

inbound queue if a logical destination has been maintained. Otherwise, the queue processing is performed under the user ID that is
currently processing a registered queue (this does not have to be the
same queue).
Therefore, a queue entry is not necessarily processed further by the
user ID that has written it into the queue; instead, it can be processed
using a different user ID. If both of these users have different rights,
a user with insufficient rights may try to process the queue entries of
another user. The resulting errors are very difficult to analyze
because, generally, they rarely occur and can seldom be reproduced.
8.4.1
Logical Destinations
We therefore urgently recommend that you create logical destinations for all inbound queues. There is little effort required, but the
benefits are great. You must perform the following steps to create a
logical destination:
1. For each queue, you should create one user, whose ID is used to
process the queue, for example, R3A_Inbound for the R3A*
queues. Start Transaction SU01, click the New button and create a
user of the type communication.
2. In Transaction SM59, check whether it has an internal connection
of the type NONE and note the name of this connection.
3. Create a new internal connection in Transaction SM59. Select Connection Type L and enter the name of the internal connection
NONE from Step 2 into the Reference Entry field on the Technical Settings tab (see Figure 8.8). Go to the Logon & Security tab
and enter the users logon information from Step 1 here (see Figure 8.9).
4. In Transaction SMQR, you can now assign the internal connection
to the corresponding queue (see Figure 8.10).
Select the corresponding queue.
Click the Registration button.
In the Queue Registration dialog box, enter the ID of the internal connection from Step 3 into the USERDEST field.
Confirm your entries.
216
Figure 8.11 shows that the logical destination R3A_INBOUND is

now assigned to all queues that start with the prefix R3A.
Repeat these steps until you have assigned a logical destination (and
therefore a separate user) to all inbound queues in Transaction
SMQR. You can now see directly from the process overview (Transaction SM50) which work processes are working on particular
inbound queues. The user also proves to be helpful in other transactions. For example, when analyzing BDoc errors (Transaction
SMW01), you can use the user ID as a filter in the User (Creator)
field if you want to select the corresponding mBDocs.
Figure 8.8 Creating a New Internal Connection of Type L (Transaction SM59)
217
8.4
Inbound Queues
Figure 8.9 Assigning a User to the New Connection (Transaction SM59)
Figure 8.10 Entering a Logical Destination (Transaction SMQR)
218
Figure 8.11 R3A Queue with Logical Destination (Transaction SMQR)
8.4.2
Message flow statistics are CRM-specific statistics that can be very

helpful when analyzing performance. You can access the message
flow statistics by using following specified path in the SAP Easy
Access menu:
Architecture and Technology Middleware Monitoring Message
Flow Display Message Flow Statistics
Alternatively, you can also use Transaction SMWMFLOW. Figure

8.12 shows the transactions start-up window. You can choose
between different statistics: the Message / Service Kernel Application Statistics and the Message / Site / Queue Statistics.
Figure 8.12 Start-Up Window of Transaction SMWMFLOW
219
8.4
Inbound Queues
Switching statistics
on/off
The writing of statistics can be switched on and off. Therefore, prior

to a performance analysis, you should check to ensure that the statistics are written. First, select Goto Activate statistics from the menu.
In the next window that opens, you can switch the kernel application
statistics and site/queue statistics on and off independently of each
other (see Figure 8.13):
Message / Service Kernel Application Statistics
When you click the Kernel application statistics button, the

window that you see in Figure 8.14 opens. To ensure that the statistics are written for the CRM Middleware, you should mark the
checkbox in the Middleware Message Hub Statistic row in the
active column. To make sure that the kernel statistics are also
actually activated, you must have already scheduled the SAP_
COLLECTOR_FOR_PERFMONITOR background job.
Figure 8.13 Activating Statistics
Figure 8.14 Activating Kernel Application Statistics
220
Message / Site / Queue Statistics
When you click the Middleware message flow statistics button

(see Figure 8.13), the window that you see in Figure 8.15 opens.
Both the Monitoring Message Flow and the Collector should be
switched on for an analysis.
Figure 8.15 Switching Site/Queue Statistics On
Workload Statistics
To display the workload statistics, click one of the two buttons Workload From Database or Last Minutes Workload in Transaction
SMWMFLOW (see Figure 8.12).

When you click the Last Minutes Workload button, the current load
of the last x minutes of the CRM instance on which you are working
is displayed. You can define how big x is yourself.
Last minutes
When you click the Workload From Database button, historical

data is displayed. You can display statistics of one instance or all
instances and choose between different time intervals.
Historical data
In both cases, the layout of the results window is identical. You

receive data about the number of BDocs per type that were processed in the time period, as well as data about the processing time,
CPU time, wait time, database time and the number of Kbytes
requested. You also receive additional information about the total
time and the average time for all time data (see Figure 8.16). Since
221
8.4
Inbound Queues
the values in milliseconds are used for the internal calculation, but
the total time is specified in seconds, this may lead to rounding variances. This occurs in particular with relatively small values or with
few statistics records.
Figure 8.16 Workload Statistics (Transaction SMWMFLOW)

Workload statistics
per service
You can receive additional detailed information about the processing

times of a particular BDoc type by selecting a row and clicking the Per
service button. The services that were called for processing a BDoc
type are then listed, and the response time, CPU time, wait time, and
database time are specified for each service (see Figure 8.17).
Figure 8.17 Workload Statistics Per Service (Transaction SMWMFLOW)

BDoc
type hierarchy
You can obtain information about the BDoc type hierarchy by selecting a BDoc type and clicking the where-used list button to the right
222
of the Per service button (see Figure 8.16). Figure 8.18 shows an
example of the BUS_TRANS_MSG BDoc type. Information about the
different times that were required for processing a BDoc or a service
is also displayed. You will also receive a range of information about
the generated BDocs.
Figure 8.18 Workload Statistics BDoc Hierarchy (Transaction SMWMFLOW)
In row , you can see that 52 mBDocs of the BUS_TRANS_MSG type

were generated with the M01 flow context. In addition to these 52
mBDocs, another 17 BUS_TRANS_MSG mBDocs were generated
with the MI0 flow context (see row ). Only the VALIDATION service was called for these 17 BDocs. You can identify the relevant predecessor and successor BDocs based on the tree structure. A total of
17 ACTIVITY_OBJECT sBDocs were processed with the SI1 flow context (see row ) and the 17 BUS_TRANS_MSG mBDocs were generated as a result. From a business point of view, this means that 17
223
8.4
Inbound Queues
activities were transferred from mobile clients to the CRM Server

during the day and validated there. The validation took 3.435 ms on
average for each mBDoc (see row ).
The 52 BUS_TRANS_MSG mBDocs with the MO1 flow context do not
have a predecessor BDoc. This means that the objects were created in
CRM Online itself and were not sent from SAP R/3 or a mobile client.
The average response time of 7.255 ms (see row ) for a BUS_TRANS_
MSG mBDoc is only provisionally significant, since a BUS_TRANS_
MSG mBDoc can contain different business objects, the processing of
which has different degrees of complexity. Underneath row , you
can see the services and functions that were called when the BUS_
TRANS_MSG mBDocs were being processed. (Note that the sequence
in the tree structure does not correspond to the actual sequence. The
actual sequence is contained in Transaction SMO8FD.)
Row contains the processing times of the Outbound Flow Service for
Mobile Clients (CRM_UPLOAD_MCA_SRV function module). In the
mobile scenario, the mBDoc is mapped to one or more sBDocs. Here,
you receive information about which sBDocs were generated when
the 52 BUS_TRANS_MSGs were processed. You can subsequently also
tell, from the names of the sBDocs, which business objects are hiding in the 52 BUS_TRANS_MSG mBDocs. When you open the tree
under row , you see that the 52 BUS_TRANS_MSG mBDocs contain
10 orders (SALESDOCGEN_O_W sBDoc), 35 activities (ACTIVITY_
OBJECT sBDoc), one service order (SRV_WRITE sBDoc), and six
opportunities (OPP_WRITE sBDoc). When you open the tree structure
under these SO1 BDocs (e.g., SALESDOCGEN_O_W in row ), you
receive information about the different services and functions that
were called when the sBDocs were being processed.
Summary
If the processing times of an object are not good enough, the statistical data in Transaction SMWMFLOW enables you to analyze exactly
which service or function and area (database, CPU) is slow. Based on
this information, you can then use a SQL or ABAP trace to determine
the database statement or coding segment where time is being lost.

To display the message flow statistics, click the Message Flow Statistics button in Transaction SMWMFLOW (see Figure 8.12). Due to
224
the decoupling of the inbound and outbound processing, the processes may run on different instances. The statistics for the inbound
and outbound processing are therefore listed separately. Within the
inbound or outbound processing, you can display the total load
(Total) or the load distribution across the individual instances. You
can also select a time interval (day or week). Figure 8.19 shows an
example of message flow statistics. The system in this example has
over five instances (us0091_Q5C_91 to us4399_Q5C_91) and daily
statistics from January 01, 2007 to January 10, 2007.
Figure 8.19 Message Flow Statistics (Transaction SMWMFLOW)
The different queue and processing times are displayed on the righthand side of the window. Depending on which tab you select, the figures for each BDoc type, site, or queue are summarized. In the Time
Profile tab, you will find information about how many BDocs were
processed in a particular period and how long the processing took. If
you select a day as the time interval, the Single Records tab will also
be displayed. The data for each BDoc is listed here individually.
To make it easier for you to understand the data, we have provided a
list of the column headers on the BDoc Type Profile tab and their
descriptions in Table 8.4 (inbound processing) and Table 8.5 (outbound processing). In Figure 8.20 (inbound processing) and Figure
225
8.4
Inbound Queues
8.21 (outbound processing), we have illustrated where the times are

measured within the Middleware.
Inbound Processing
Inbound Queues
R/3
R3A*
BAPI
R/3 Adapter
Mapping
Groupware
ISP*
XML
Groupware
Adapter
Mapping
Mobile
Adapter
Mapping
mBDoc
CRM SITE*
CRM SITE*
Mobile
Clients
sBDoc
Validation
Validation Service
CRM
Application
Online
DB
CRM SITE*
Figure 8.20 Times for Inbound Processing (Transaction SMWMFLOW)

Times for inbound
processing
Column title Description

Synch BDoc
Name of synchronization BDoc type
Messaging
BDoc type
Name of messaging BDoc type
Processed
Number of BDocs of this type processed in total
Queued
Number of BDocs of this type in the inbound queues
Queue(s)
Total wait time in seconds in the inbound queues
Avg. (s)
Average wait time in seconds in the inbound queues
Inbnd. (ms)
Total processing time in milliseconds in the inbound adapter
Avg. (ms)
Average processing time in milliseconds in the inbound adapter
Mappg. (ms)
Total processing time in milliseconds for mapping the BDocs
Avg. (ms)
Average processing time in milliseconds for mapping the BDocs
Mflow. (ms)
Total processing time in milliseconds in the message flow
Avg. (ms)
Average processing time in milliseconds in the message flow
Table 8.4 Descriptions of Columns in Inbound Processing (Transaction

SMWMFLOW)
226

Proc. (ms)
Total processing time in milliseconds (without the wait time in

the inbound queue)
Avg. (ms)
Average processing time in milliseconds (without the wait time in

the inbound queue)
Table 8.4 Descriptions of Columns in Inbound Processing (Transaction SMWMFLOW) (cont.)
Outbound Processing
CSA*
Outbound Queues
Replication
Service
R/3
R3A*
BAPI
R/3 Adapter
Groupware
ISP*
XML
Groupware
Adapter
sBDoc
Mobile Bridge
CRM SITE*
CRM SITE*
Mobile
Clients
Mobile
Adapter
CDB
Service
CDB
CRM SITE*
sBDoc
R&R Service
Figure 8.21 Times for Outbound Processing (Transaction SMWMFLOW)

Times for
outbound
processing

BDoc type
Name of BDoc type
Processed
Number of BDocs of this type processed in total
Queued
Number of BDocs of this type in the inbound queues
Queue(s)
Total wait time in seconds in the inbound queues
Avg. (s)
Average wait time in seconds in the inbound queues
MFlow.
(ms)
Total processing time in milliseconds in the message flow
Avg. (ms)
Average time in milliseconds in the message flow
Table 8.5 Descriptions of Columns in Outbound Processing (Transaction

SMWMFLOW)
227
8.4
Inbound Queues

SFlow. (ms)
Total processing time in milliseconds in the synchronization flow
Avg. (ms)
Average time in milliseconds in the synchronization flow
SFlow. cnt.
Number of BDocs of this type in the synchronization flow
Avg.
Number of BDocs in the synchronization flow divided by the

number of processed BDocs
Proc. time
(ms)
Total processing time in milliseconds (without the wait time in

the inbound queue)
Avg. (ms)
Average time in milliseconds (without the wait time in the

inbound queue)
Table 8.5 Descriptions of Columns in Outbound Processing (Transaction SMWMFLOW) (cont.)
8.4.3
Setting trace levels
CRM Middleware Trace
The CRM Middleware Trace enables you to obtain additional information that is written into the Middleware during processing. The CRM
system not only allows you to switch the writing of the Middleware
trace on and off, it also enables you to define the area where the trace
is written and its granularity. For example, you can specify that the
trace is restricted to errors and warnings in the message flow (Trace
level: Warning), whereas you want all information (Trace level:
Detail Level 2) to be written during the generation (see Figure 8.22).
To maintain a trace level, go to Architecture and Technology Middleware Monitoring Message Flow Set up Middleware Trace
from the SAP Easy Access menu or by using Transaction SMWTAD.

Trace levels
The following trace levels are available:
Level 0: Error
Only serious errors are reported.
Level 1: Warnings
Only situations that can lead to an error are reported.
Level 2: Service Flow

A note is made of all services that are processed in the Middle-
ware.
Level 3: Detail Level 1

Additional information is written about the executed programs
and modules.
228
8.4
Level 4: Detail Level 2
Program-specific information is written.
Figure 8.22 Settings for Middleware Trace (Transaction SMWTAD)
The standard setting in the live system is Level 1. Levels 3 and 4 are
intended for developers.
For performance and disk space reasons, you should delete old traces
on a regular basis. SAP provides the SMO6_REORG2 report for reorganization purposes, which you can also use (among other things) to
delete traces.
Chapter 12, Reorganization, contains information about this report,
and reorganization in general.
You can access the trace itself by using the following path in the SAP
Easy Access menu:
Architecture and Technology Middleware Monitoring Message
Flow Display Middleware Trace
Alternatively, you can also use Transaction SMWT. Figure 8.23

shows the selection screen that appears when you start the transaction.
229
Displaying a trace
Inbound Queues
Figure 8.23 Selection Screen for Middleware Trace (Transaction SMWT)
In accordance with the selection criteria you enter, a list of the available traces in the system is displayed. You can display the trace by
double-clicking the corresponding row (see Figure 8.24).
Figure 8.24 Middleware Trace
230
Dependencies Between Inbound Queues
The trace itself contains a range of information: in addition to the

Date, Time, Environment and trace level (Level), the Text Field and
Trace GUID in particular are displayed. Text Field contains the trace
message (maximum of 250 characters) and Trace GUID contains the
GUID of the trace object. If youre searching for the trace messages
for a particular BDoc, for example, you can search according to the
GUID of the BDoc in the Trace GUID field. Alternatively, you can
enter the BDoc GUID as your selection criterion in the GUID1 field;
only the traces that contain messages for the BDoc will subsequently
be displayed (see Figure 8.23). In the case of BDocs in particular, you
can also jump directly to the trace from Transaction SMW01.
To do this, select the BDoc in Transaction SMW01 and click the Middleware Trace button, which is highlighted (see Figure 8.25).
Figure 8.25 Jumping to the Middleware Trace from Transaction SMW01
8.5
You cannot always process CSA inbound queues in parallel on an

unrestricted basis. Dependencies may occur between individual
queues for some object types, which means that the processing for a
231
8.5
Inbound Queues
queue must wait until one or more entries of another queue are processed. The following example illustrates this problem, based on a
download of business partner relationships from SAP R/3.
Example
A request for all business partner relationships is started in a CRM system.

The number of R3AR_BUPA* queues is limited to three and there is no
limit on the number of CSABUPA* queues. A BUPA_REL BDoc from R/3
contains all relationships of a business partner (to simplify matters, lets
assume that all business partners have four relationships). If an RFC record
is being processed in a R3AR queue, the individual relationships are written into different CSA queues (depending on the business partner GUID in
the relationship). Therefore, this results in five CSA queue entries (and
consequently the corresponding CSA queues), even though only one
mBDoc is generated. If the mBDoc is being processed, you see in Transaction SMQ2 that the five queues have the running status; while in Transaction SM50, you see that only one work process is busy.
The BDoc can only be processed if all CSA queue entries are in the first
position of the queue in question, since otherwise, the queue entries
would be superseded.
If the number of CSA queues is not restricted, they usually have one entry.
Two or more entries will only appear in a CSABUPA queue if two business
partners have a relationship to one another or to a common third party.
Figure 8.26 shows that both business partners 1111 and 2222 have a relationship to business partner 4112. If both entries in the R3AR_BUPA1111
and R3AR_BUPA2222 queues are processed, two entries are written into
the CSABUPA4112 queue. Only one entry is written into all other
CSABUPA* queues. The BUPA_REL BDocs in the CSA queues for business
partners GP1111 and GP3333 can be processed immediately and in parallel, since the relevant entries are situated in the first position in the CSA
queues. The BUPA_REL BDoc for business partner GP2222 cannot be processed immediately, because the GP2222-GP4112 relationship is not in
the first position in the CSABUPA4112 queue. All queues that have a business partner relationship to GP2222 as the first entry (CSABUPA412*) get
the waiting status, since they can only be processed after the GP1111GP4112 business partner relationship has been processed in the
CSABUPA4112 queue.
If there is no limit to the number of CSA queues (as described in the

example), situations only rarely occur where queues have to wait for
each other, or so many queues are created that this does not affect
the processing speed. It is a different situation if the number of CSA
232
queues has been severely limited. The CSA queues generally have
more entries and the number of queues waiting for each other
increases, as explained in the following example.
Status
GP1111
GP4111
running
GP1111
GP4112
running
CSABUPA4113
GP1111
GP4113
running
CSABUPA4114
GP1111
GP4114
running
CSABUPA4125
GP2222
GP4125
waiting
CSABUPA4126
GP2222
GP4126
waiting
CSABUPA4127
GP2222
GP4127
waiting
CSABUPA4137
GP3333
GP4137
running
CSABUPA4138
GP3333
GP4138
running
CSABUPA4139
GP3333
GP4139
running
CSABUPA4130
GP3333
GP4130
running
CSABUPA4111
R3AR_BUPA1111
GP1111
CSABUPA4112
GP2222
GP4112
GP4111
GP4112
GP4113
GP4114
R3AR_BUPA2222
GP2222
GP4112
GP4125
GP4126
GP4127
R3AR_BUPA3333
GP3333
GP4137
GP4138
GP4139
GP4130
Figure 8.26 Dependencies Between CSA Queue Entries
Example
The prerequisites in this example correspond to those from the last example, however, the only difference in this case is that the number of
CSABUPA* queues has been limited to 10.
Figure 8.27 shows that the relationship between GP2222 and GP4127
and the relationship between GP3333 and GP4137 are written into the
same queue due to the different queue naming.
Consequently, the BUPA_REL BDoc of business partner 3333 can only be
processed if the GP2222-GP4127 entry of queue CSABUPA4127 (i.e., the
BUPA_REL BDoc of business partner 2222) has been processed. However,
this entry can only be processed if the first entry GP1111-GP4112 of
queue CSABUPA4112 (i.e., the BUPA_REL BDoc of business partner
1111) has been processed. In other words, this means that the BUPA_REL
BDocs are processed sequentially, even though 10 CSA queues exist.
233
8.5
Inbound Queues
Status
R3AR_BUPA1111
GP1111
GP4111
running
GP1111
GP4112
running
CSABUPA4113
GP1111
GP4113
running
CSABUPA4114
GP1111
GP4114
running
CSABUPA4125
GP2222
GP4125
waiting
CSABUPA4126
GP2222
GP4126
waiting
GP2222
GP4127
waiting
CSABUPA4138
GP3333
GP4138
waiting
CSABUPA4139
GP3333
GP4139
waiting
CSABUPA4130
GP3333
GP4130
waiting
CSABUPA4111
GP1111
GP4111
GP4112
CSABUPA4112
GP2222
GP4112
GP4113
GP4114
R3AR_BUPA2222
GP2222
GP4112
GP4125
GP4126
GP4127
CSABUPA4127
R3AR_BUPA3333
GP3333
GP3333
GP4137
GP4137
GP4138
GP4139
GP4130
Figure 8.27 Dependencies with Limited Number of CSA Queues
234
Index
.NET Connector 45
A
AC extract 119, 281
bulk extract 283
unfiltered extract 283
AC_EXTRACT queue 111
Adapter 63
Adapter framework 63
Adapter object 64
activating and deactivating 68
assignment to a BDoc type 65
block size 65
business object 64
condition object 64
customizing object 64
filter settings 72
initial flow context 69
mapping module CRM R/3 73
object class 67
parent objects 73
tables/structures 70
Administration console 31, 90
improvements 365
wizard 98
Analysis roadmap
CPU bottleneck analysis 345, 397
inbound queue processing 336, 337,
395
R&R optimization 341
work process analysis 334, 395
ARFCRDATA, Table 47, 50
ARFCRSTATE, Table 46, 47
ARFCSDATA, Table 46, 47, 49
ARFCSSTATE, Table 46, 47
B
Background RFC 367
BAPI structure 25
BAPIMTCS 171
BDoc 37, 38
assignment of segment and database
143
class 135
error status 148

final Status 149
instance 140
interim status 148
release 146
robust data storage 152
sendbits 147
standard field 147
static WHERE clause 144
status 147
task type 147
type 140
BDoc instance 40, 140
BDoc link
adjacent functionality 316
avoiding 315
reorganization 314
BDoc merge 364
BDoc message 39
BDoc Modeler 140
BDoc release, WHERE clause 146
BDoc statistics, reorganization 309
BDoc store 75, 79
BDoc type 26, 39
lock 146
bgRFC see Background RFC 367
Block size 278
Bulk extract 283
Bulk message 123
Business document 38
Business object, adapter object 64
C
CDB 31, 107
CDB service 31, 107, 108
Checkbox structure 175
Classic part see mBDoc 137
Communication monitor, reorganization 310
Condition object, adapter object 64
Confirmation message 122
ConnTrans 32, 74
transfer duration 272
Consolidated database see CDB 31
CP_CODEPAGE 302
401
Index
Creating a database table 172

Creating an object class 181
CRM_MAX_QUEUE_NUMBER_DELT
A, parameter 204, 207
CRMPAROLTP, Table
CRM_XML_BACKGROUND_PROCES
SING_ON 299
CRMPAROLTP, table
number of inbound queues 204
CRMQNAMES, Tabelle
FLDOFFSET, Feld 202
CRMQNAMES, Table 61, 204
LENGTH, field 202
CRMRFCPAR, Table
XML control 297
CRMSUBTAB, Table 67, 171
CRS_FIRST_DOWNLOAD_TRIGGER
171
CSA queue 29, 101
Current state message 122
Customizing object, adapter object
64
D
Data collector 253, 281, 295
reorganization 312
Data Integrity Manager 134
Data transfer
delta 63
initial 63
DB statistics 261
DBSTATC, Table 262
Default pool 276
Deletion message 123
avoiding 247
Delta data transfer 63
Destination
excluding 127
parameters 127
registering and deregistering 127
DIMa see Data Integrity Manager
134
Disk subsystem 274
Distribution
bulk 238
intelligent 237
intelligent, switching to bulk 242
intelligent, without filter criteria 241
402
Distribution model 31
Dynamic mapping 76
E
Error Handler 78, 81, 82
EXEMODE, Parameter 54
Extended Markup Language see
XML 297
Extension part see mBDoc 137
EXTRACT queue 111
EXTRACTBLK queue 111
Extractor 171
F
Flow 37, 40
context 40, 179
definition 42
G
Generated mobile inbound adapter
74
Generic mobile inbound adapter 74
GNRWB, Transaction 304
Groupware adapter
GWA_01 156
GWA_02 156
Groupware connector see Groupware integration 158
Groupware integration
analysis 164
client-client scenario 155
data queue, primary 160
data queue, secondary 160
folder, private 156
folder, public 156
groupware adapter 156
groupware connector 158
groupware connector proxy 158
MapBox 157
MapBox, log files 164
MapBox, RFC destination 158
MBMANDTSTORE, table 162
overview 155
payload interface 158
server-server scenario 155
system queues 160, 164
Index
trace of internal SyncPoint 165

trace of payload interface 166
userlist.xml 162
GWI see Groupware integration
155
H
Hardware Bottleneck 209
I
Inbound adapter 26, 37, 63, 69
Inbound processing 37
data from R/3 27
mobile client data 35
Inbound queue 26, 33, 37, 44, 48
dependencies 231
deregistering 54
details 59
entries 60
of mobile client 32
overview 56
parameters 54
reduce number of queues 200
registering 52
slow processing 199
status 58
Inbound queue name
data from mobile clients 61
data from SAP R/3 61
for CSA queues 101
Inbound queue scheduler 45, 51
activating 52
performance 199
status 51
Inbound scheduler 51
Index fragmentation 262
Index quality 262
Initial data transfer 63
Initial load 183
Integration model
message exchange 136
synchronization 136
Interlinkage 98, 248
J
Java Connector (JCo) 45
K
KEEP pool 276
Kernel application statistics 220
L
Logical Unit of Work 47
Lookup table 31, 90, 252
LUW 47
M
MapBox see Groupware integration
157
Mapping 75
BAPI container in mBDoc 180
dynamic 76
mBDoc to sBDoc 31
static 75, 139
Mapping function module 26
Mapping method 34
Mass change 188
planned 326
unplanned 326, 327
MAX_PACKAGE_SIZE, parameter
278
MAXTIME, Parameter 54, 353
mBDoc 26, 38, 135
classic part 137
creating the classic part 177
creating the extension part 174
extension part 137
MBMANDTSTORE, table 162
Message flow statistics 224
kernel application statistics 220
Middleware message flow statistics
221
switching on/off 220
Messaging BDoc see mBDoc 26, 38
Messaging flow 101
Middleware message flow statistics
221
Middleware trace 228
displaying a trace 229
Reorganization 309
setting trace levels 228
Mobile adapter 107
Mobile application BDoc 136
403
Index
Mobile bridge 30, 90, 102, 104

Mobile inbound adapter 34
generated 74
generic 74
Mobile outbound adapter 31, 122
N
Naming for queues
advantages/disadvantages 206
Neighbour functionality 315
NRETRY, Parameter 54
O
Online database 26, 37
Outbound adapters 69, 90, 102, 104
Outbound processing 89
for mobile clients 32
for R/3 29
Outbound qRFC with recipient list
269
benefits 270
Outbound queue 29, 31, 48, 89, 124
displaying an overview 128
displaying details 130
displaying entries 131
in R/3 25
of mobile client 33
status 129
Outbound queue name
additonal ones 134
data to mobile clients 132
data to R/3 132
Outbound Scheduler 125
P
Parallel processing, optimizing the
middleware 318
Payload interface see Groupware
integration 158
Performance analysis 215
SMWMFLOW 219
SMWT 228
Publication 93
404
Q
QIN Scheduler see Inbound queue
scheduler 51
QOUT Scheduler 125
activating 126
status 126
QREFTID, Table 49
qRFC 45, 48, 125
monitor for inbound queues 56
qRFC Monitor
for Outbound Queues 128
qRFC monitor for inbound queues 45
Query BDoc see Mobile application
BDoc 136
Queue naming
changing the naming for queues 201
Queue, stopping 257
Queued RFC see qRFC 45
R
R&R 285
definition 235
DEPENDENCY queue 360
distribution-relevant fields 238
internal optimization 264
new improvements 360
new queue framework 360
optimizing a mass change 267
parallel processing of queues 361
parallelizing queue processing 256
processing queues in blocks 362
queues 236
realignment 235
replication 235
replication wrapper 238
R&R queue 111
displaying 111
starting and stopping 114
status 113
R&R queue demon 112
starting and stopping 113
status 113
R&R queue framework 110
R&R service 31, 107
R/3 outbound adapter 29
R3AC1, Transaction 64, 182
R3AC3, Transaction 64, 180
Index
R3AC5, Transaction 64
R3AC6, Transaction
controlling the reorganization process 310
queue parallelization 256
R3AM1, Transaction 183
R3AR2, Transaction
one-time request 314
R3AS, Transaction 183
REALIGN queue 111
Realignment 31, 109
Rejection message 78, 123
Remote Function Call see RFC 45
Reorganization
BDoc links 314
BDoc messages 308
BDoc statistics 309
data collector 312
key generation 309
middleware trace 309
of data for sites that cannot be activated 313
request 314
SAP_MW_REORG, variant 308
SMO6_REORG 308
SMO6_REORG2 308
SMW3* tables 308
standard variant 308
statistics of the CommStation sessions
310
subscription agent 313
Replication 31, 108, 114
bulk 115
intelligent 116
Replication & Realignment 287
Replication & Realignment see R&R
Replication & Realignment service
see R&R service 31
Replication model 90
optimization 239, 248
optimizing interlinkages 248
optimizing the bulk publications 240
Replication object 90
Replication object type 91, 92
bulk 96
Dependent 97
Intelligent 96
Simple Bulk (MESG) 95
Simple Intelligent (MESG) 96
Simple Intelligent (SYNC) 96
Replication service 29, 102

mBDoc 103
sBDoc 114
Replication wrapper
mBDoc 103
sBDoc 114
RFC 45
RFC libraries 45
RFC server group 209, 257
assign 213
create 210
optimal number 325
parameters 211
profile parameters 212
RFC Software Development Kit 45
RSANAORA, Report 263
RSRLDREL, program 314
RSTRFCQD, Report 355
RSTRFCQDS, Report 355
RZ12, Transaction 210
S
SBDM, Transaction 74, 77, 140, 177
sBDoc 30, 38, 135
block size 278
structure 136
Scheduler, tRFC 46
SDIMA, Transaction 134
SDK, (RFC) Software Development Kit
45
SE11, Transaction 172
Segment, field assignment 136
Service, generating 182
Setting up a logical destination 216
Site 93
deactivating 244
mass deactivation 246
Site type 94
site decativation not supported 313
Sizing 277
SM50, Transaction
occupancy of the work processes 332
SM59, Transaction 54, 126
setting up a logical destination 216
SMO 275
SMO8FD, Transaction 42, 79
SMO9_KYTBL, Table
reorganization 309
405
Index
SMOE_BULK_SITE_ACTIVATION,
Report 285
SMOEAC, Transaction 61, 90, 132
AC extract 281
XML optimization 303
SMOECK, Transaction 241
SMOEGENDET, Table 100
SMOEGENHEA, Table 100
SMOEGENLOG, Table
reorganization 313
SMOEJOBID, Table
reorganization 310
SMOFFILTAB, Table 72
SMOFINICON, Table 69
SMOFOBJCLA, Table 68
SMOFOBJECT, Table 64, 65, 68
SMOFOBJPAR, Table 73
SMOFPARSFA, Table 301, 310
deactivating mBDoc links 315
processing R&R queues in bloks 362
SMOFPARSFA, table
block size 278
SMOFQFIND, Table 204
SMOFQNAMES, Table 132
SMOFSUBTAB, Table 73
SMOFUPLMAP, Table 73
SMOGGEN, Transaction 182, 244
SMOHILTP, Table 99
SMOHJOBQ, Table 259
SMOHLUBULK, Table 96, 238
SMOHMSGQ, Table 111, 259
optimizing the access path 259
SMOHMSGST, Table 259
SMOHPUBL, Table 93
SMOHQTAB, Table 61, 132
SMOHQUEUE, Transaction 111, 113,
114, 236, 243
AC extract 282
block size 278
Stop queue 353
SMOHREPOBJ, Table 92
SMOHSGQST, Table 111
SMOHSITEID, Table 94
SMOHSITEQ, Table 111, 259
SMOHSUBSCR, Table 95
SMOHSUBSIT, Table 95
SMOJDC, Transaction 255
SMOJDCPROC, Table
reorganization 312
SMQ1, Transaction 128
406
SMQ2, Transaction 56
SMQR, Transaction 51
changing RFC server group 213
MAXTIME 353
setting up a logical destination 216
SMQS, Transaction 125
SMW01, Transaction 83, 316
jumping to the Middleware trace
231
SMW01, transaction
DEBUGMODE 152
SMW1SPRVDR, Table 313
SMW3*, Table
reorganization 308
SMW3BDOCIF, Table 43, 79, 179
SMW3FDBDOC, Table 44, 105
SMW3FDBDOC, Transaction 44, 105
SMW3FDCUST, Table 44
SMW3FDCUST, Transaction 44
SMW3FDIF, Transaction 44, 76, 79,
179
SMW3FDSTD, Table 44
SMW3FDSTD, Transaction 44
SMWMCOMM, Transaction 310
SMWMFLOW, Transaction 219
message flow statistics 224
workload statistics 221
SMWMSESSHT, Table 310
SMWMSESSIN, Table 310
SMWT, Transaction 229, 309
SMWT_TRC, Table
reorganization 309
SMWTAD, Transaction 228
SPRO, Transaction 82
ST03N, Transaction 309, 310
ST06, Transaction
Idle time 342
Load average 344
Static mapping 75
Storage Area Network 274
Storage quality 276
SUBCHECK queue 111
Subscription 94
changing the assignment of sites 120
Subscription agent 99
reorganization 313
Subscription generators 99
Synchronization 63
Synchronization BDoc see sBDoc
38
Index
between CRM Server and a mobile client 300

between R/3 and CRM 297
code page conversion 300
data exchange between R/3 and CRM
297
data exchange with mobile clients
300
decoupling the application from the
conversion 299
performance increase for mobile client
data exchange 300
restrictions regarding optimization
305
synchronous RFC 298
Synchronization flow 90, 107

System landscape 187
heterogeneous 297
homogeneous 297
inhomogeneous 297
System monitoring 330
System Optimization Services 275
T
TDELAY, Parameter 54
TID 47
Transaction Identifier 47
Transactional Remote Function Call
see tRFC 45
Transactional RFC see tRFC
tRFC 45, 46
TRFCQDATA, Table 50
TRFCQIN, Table 50
TRFCQOUT, Table 49, 50
TRFCQSTATE, Table 50
TRFCRSTATE, Table 50
TRFCSDATA, Table 50
TRFCSSTATE, Table 50
TSMW3_STAT, table 147
Z
ZAP message 123, 284
U
Upload 63
USERDEST, Parameter 54
V
Validation 37, 79, 179
data from R/3 27
mobile client data 35
Validation service 26, 79
W
WHERE clause see BDoc 144
Wizard 98
Workload statistics 221
switching on/off 220
X
XML 297
407

Queues Optimization

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Queues Optimization

Uploaded by

Copyright:

Available Formats

121.book Seite 197 Freitag, 3.

August 2007 2:19 14

Data sent to CRM is usually written into the inbound queues

If we consider the different systems that can be connected to CRM,

Figure 8.1 Inbound Queues in CRM Middleware

121.book Seite 198 Freitag, 3. August 2007 2:19 14

This delta supply of data works well in normal operations. However,

Figure 8.2 RFC in CRM Middleware

Chapter 2, Inbound Processing and Validation, contains a detailed

121.book Seite 199 Freitag, 3. August 2007 2:19 14

The Number of Inbound Queues Is Too High

So many inbound queues are created that the inbound queue

The inbound processing of individual data records is slow (see

Dependencies exist between inbound queues (see Section 8.5).

In the following sections, we will describe the different causes of

The Number of Inbound Queues Is Too High

121.book Seite 200 Freitag, 3. August 2007 2:19 14

The company, BodeGolzeSchrder Inc. (hereafter abbreviated to BGS),

Reduce the single records in R/3

Change the naming for queues

Use R/3 parameter CRM_MAX_QUEUE_NUMBER_DELTA (which

The first approach involves reducing the single records generated by

121.book Seite 201 Freitag, 3. August 2007 2:19 14

The Number of Inbound Queues Is Too High

ASSIGNMENT report would only create 600, rather than 300,000,

121.book Seite 202 Freitag, 3. August 2007 2:19 14

By changing the naming for queues, several business partners are

1. Setting: Standard Naming

R/3 Customer Number

CRM Queue Name

121.book Seite 203 Freitag, 3. August 2007 2:19 14

The Number of Inbound Queues Is Too High

2. Setting: FLDOFFSET = 9 and LENGTH = 1

CRM Queue Name

CRM Queue Name

CRM Queue Name

121.book Seite 204 Freitag, 3. August 2007 2:19 14

Using the CRM_

The third approach for reducing the number of inbound queues

121.book Seite 205 Freitag, 3. August 2007 2:19 14

The Number of Inbound Queues Is Too High

Proceed as follows to limit the number of R3AD queues (see Figure

121.book Seite 206 Freitag, 3. August 2007 2:19 14

Figure 8.5 Table CRMPAROLTP in SAP R/3

Max. Number of Queues

Table 8.1 Maximum Numbers of Inbound Queues Depending on the Type of

The actual number of queues is influenced to a certain extent by the

121.book Seite 207 Freitag, 3. August 2007 2:19 14

The Number of Inbound Queues Is Too High

To be able to identify a customers country of origin immediately based on

You could say that the CRM_MAX_QUEUE_NUMBER_DELTA

We recommend that you first deregister the outbound queues in R/3

Reducing SAP R/3

121.book Seite 208 Freitag, 3. August 2007 2:19 14

The Number of Inbound Queues Is Too Low

In Section 8.1, we described how you could limit the number of

In some circumstances, you can solve the problem described in the

121.book Seite 209 Freitag, 3. August 2007 2:19 14

choose the correct balance between the number of inbound queues

RFC server group

121.book Seite 210 Freitag, 3. August 2007 2:19 14

ment to a group. The other two groups, Queue_Scheduler and